Profiler analysis from Cosimulation (Matlab+GT(S-function))

Using Matlab profiler tool to analyse a Cosimulation (Matlab+GT-Suite) result conducted on Matlab, I discover the following computational cost distribution associated to the Simulation Phase:

1. Update 34.80%
2. Outputs.Major 54.60%,
3. SolverPhase 2.30%
4. Outputs.ParameterChangeEvent 0 %

As expected, 90% of the Update group's effort is due to the S-Function which links Matlab to GT-software and some extra model's reference updates.

However, the high computational cost corresponding to the Output phase is something I did not expect and I can not explain. A list of the Ouput.Major computational cost distribution can be found below:

1. RateTransition.Output.Major (Number of calls Nx248)
2. AtomicSubsystem.Output.Major (Number of calls N)
3. ModelReference.Output.Major (Number of calls N)
4. EnabledSubsystem.Output.Major (Number of calls N)

Can somebody help me to understand the reason on this distribution in order to improve the simulation phase?

Kind Regards